The most common mistake store owners make with Google Ads bidding is switching strategies too early, too aggressively, or without understanding what triggers the learning period. Campaigns that were performing well go quiet for two weeks. ROAS crashes. Spend drops to almost nothing. Then someone makes another change trying to fix it, and the cycle repeats.
This guide covers how Google’s Smart Bidding strategies actually work, when you have enough data to transition between them, how to make the transition without destabilizing your campaign, and how to diagnose the specific failure modes that appear when bidding goes wrong.
How Smart Bidding Actually Works
Smart Bidding strategies — Maximize Conversions, Maximize Conversion Value, Target CPA, and Target ROAS — use Google’s machine learning to set bids at auction time. Instead of you setting a fixed bid, Google estimates the probability of a conversion for each individual auction and sets the bid accordingly.
This bid calculation uses signals from the auction: the user’s device, location, time of day, search query, browsing history, and hundreds of other factors. The more conversion data your campaign has, the better Google can estimate conversion probability and set bids accurately.
This is why conversion data volume matters so much for Smart Bidding. With few conversions, the probability estimates are uncertain and bids are inconsistent. With more conversions, the model becomes more accurate and bids stabilize.
The Four Bidding Strategies and When Each Is Right
Maximize Conversions — Google sets bids to get as many conversions as possible within your daily budget. No ROAS or CPA constraint. This is the right strategy when you are in the data-gathering phase: your campaign needs conversion history before efficiency constraints can be set meaningfully.
Use this when: new campaigns, account restructures, campaigns with fewer than 30 conversions in the last 30 days, or when you have been given a budget to spend and volume is the priority.
Maximize Conversion Value — Google maximizes the total conversion value (revenue) rather than conversion count. This is appropriate when your conversions have variable values — different products at different price points — and you want Google to optimize toward higher-value orders rather than just more orders.
Use this when: you have conversion value data (purchase revenue tracked correctly), your product catalog has meaningful price variation, and you want Google to favor higher-value transactions.
Target CPA (tCPA) — Google sets bids to achieve your target cost per acquisition. You set a number; Google aims for it. Use this for lead generation or when a consistent CPA ceiling matters more than volume.
For ecommerce, tCPA is less commonly the right tool than tROAS because product value varies. tCPA treats a $15 sale and a $300 sale as equivalent conversions.
Target ROAS (tROAS) — Google sets bids to achieve your target return on ad spend. You set a ratio (e.g. 400% = $4 revenue for every $1 spent); Google aims for it. This is the target state for most mature ecommerce campaigns because it constrains efficiency relative to revenue value.
Use this when: you have consistent conversion volume, conversion value tracking is accurate, and you want to optimize for profitability rather than raw volume.
The Conversion Threshold Question
The frequently asked question — “how many conversions do I need before switching to tROAS?” — has a real answer, and it is not a single number.
Google’s official recommendation is 50 conversions in a 30-day window before adding a ROAS target. This is a reasonable minimum. Below 50, the model does not have enough data to estimate conversion probability accurately, and adding a ROAS constraint on top of an uncertain model leads to the campaign either cutting spend aggressively (because it is uncertain which auctions will hit your target) or overshooting in ways that are hard to predict.
The practical reality: accounts with 15-30 conversions per month can sometimes run tROAS successfully if the conversion data is very consistent and the target is set conservatively. But the risk of instability is much higher below the 50 threshold.
More important than the count is the consistency. 50 conversions spread evenly across 30 days supports better modeling than 50 conversions clustered around one sale event.
If you are below the conversion threshold, Maximize Conversions (without a CPA constraint) is the right strategy. The goal at this stage is to build conversion history, not optimize efficiency. Constraining an under-data campaign is counterproductive.
When to Transition from Maximize Conversions to Target ROAS
The right time to transition is when:
- You have 50+ conversions in the last 30 days from the specific campaign you are transitioning
- The campaign has been stable for at least 2 weeks without major structural changes
- Your conversion tracking is verified as accurate (correct events, no duplicate firing, purchase revenue tracked correctly)
- You have a realistic target ROAS based on actual historical performance
On that last point: the most common transition mistake is setting an aspirational tROAS rather than a realistic one. If your campaign has been running at 350% ROAS on Maximize Conversions, setting a 600% tROAS target on day one will cause the campaign to cut spend dramatically — it will only enter auctions where it estimates achieving 600% ROAS, which are rare, so volume collapses.
How to set the initial tROAS target: Look at your last 30-day average ROAS on the current strategy. Set your tROAS target at roughly 85-90% of that number to start. This gives the campaign room to operate while nudging efficiency upward. If the campaign maintains volume and hits the target for 2 weeks, raise the target by 10-15% and watch for 2 more weeks.
Tightening the target gradually is far less disruptive than setting an aggressive target immediately.
The Learning Period: What It Is and Why It Matters
Any time you make a significant change to a Smart Bidding campaign, Google flags it as being in a learning period. During this period:
- Bid estimates are less accurate than normal
- Performance is typically below the campaign’s baseline
- Google’s documentation says this lasts 1-2 weeks, but it can extend to 3-4 weeks for campaigns with lower conversion volume
Changes that trigger a learning period:
- Switching bid strategies
- Changing a tROAS or tCPA target by more than 15-20%
- Significantly changing the campaign’s budget
- Adding or removing conversion actions
- Significant structural changes (new ad groups, major keyword additions)
The key implication: do not make multiple significant changes in rapid succession. Each change resets the learning period. A campaign that keeps getting changed never fully exits the learning phase and never reaches its optimization potential.
If you need to make multiple changes, sequence them with at least 2 weeks between significant adjustments. Make the most important change first, let the campaign stabilize, then make the next one.
Why Your tROAS Is Cutting Spend Too Aggressively
If you switched to tROAS and your campaign spend dropped significantly, the most likely causes are:
Target is set too high relative to actual achievable ROAS. The campaign is entering fewer auctions because few auctions are predicted to hit your target. Fix: lower the tROAS target to closer to your historical average and let the campaign rebuild volume before tightening again.
Conversion volume dropped after the transition. Sometimes switching strategies coincides with a natural drop in market demand (seasonality, competitor activity) or a conversion tracking issue. Separate the bidding strategy effect from external factors before concluding the strategy change caused the drop.
Campaign is still in the learning period. Spend instability in the first 1-2 weeks after a strategy change is normal. If it has been more than 3 weeks and spend is still suppressed, investigate further.
Budget is too low relative to the target ROAS. If your daily budget is $50 and your tROAS is 800%, the campaign may find very few auctions it can enter confidently. A budget that severely constrains the number of auctions available can cause tROAS campaigns to underdeliver. Increase the budget or lower the target.
Recovering a Campaign That Crashed After a Bid Strategy Change
If a campaign has been destabilized by a bid strategy change and performance is poor, the recovery sequence matters.
Do not immediately switch back to Maximize Conversions. Switching strategies again resets the learning period for a third time and digs the hole deeper.
Instead:
-
If you switched to tROAS and set an aggressive target, lower the target to match historical ROAS. Give the campaign 2 weeks at the lower target before evaluating.
-
If the campaign is spending but ROAS is poor, verify your conversion tracking has not changed. A conversion tracking issue that appeared around the same time as the strategy change is a common cause of post-transition performance confusion.
-
If the campaign has nearly stopped spending, check whether the budget has become the bottleneck. A tROAS constraint on a small budget is often the cause.
-
If after 3-4 weeks the campaign is still underperforming significantly, reverting to Maximize Conversions to rebuild conversion data is a valid option — but accept that you will lose 2 more weeks to another learning period in the process.
The Bid Strategy Lifecycle for a New Ecommerce Campaign
For a new campaign starting from zero:
Months 1-2: Maximize Conversions, no CPA or ROAS constraint. Budget at a level that allows 2-3 conversions per day minimum. Goal: accumulate conversion history. Do not evaluate efficiency during this phase.
Month 2-3 (if 50+ conversions accumulated): Introduce tROAS at a conservative target — 80-90% of your actual achieved ROAS. Monitor for 2 weeks. If volume holds and ROAS hits target, raise target by 10%.
Ongoing: Gradually tighten tROAS target in 10-15% increments with 2-week observation periods between adjustments. The ceiling is where the campaign maintains acceptable volume. If volume collapses at a certain target, the campaign has reached its efficiency limit at current scale — growth requires either accepting a lower target or improving conversion rate and feed quality to expand the addressable auction pool.
This lifecycle applies to Shopping campaigns, PMax, and Search campaigns. The thresholds and timelines are similar across campaign types, though PMax typically needs more conversion data before targets can be set tightly.
One Rule That Prevents Most Bidding Problems
Make one significant change at a time, and wait for the learning period to complete before evaluating results.
Most bidding crises come from making a change, seeing it not work immediately, making another change to fix it, making a third change when that does not work either — and ending up 6 weeks later with a campaign that has been in a perpetual learning period and a performance record that no longer reflects any coherent strategy.
Change something. Wait two weeks. Evaluate. Then change the next thing. It is slower but it is the only way to know what is actually working.
Related Posts
Incrementality Testing: How to Know If Your Google Ads Are Actually Driving Sales
Google Ads for Seasonal Businesses: How to Handle Budget, Campaigns, and Bidding Around Peaks
Google Ads for International Ecommerce: Multi-Country Campaigns, Currency, and Feed Setup
Need Help With Your Google Ads?
I help e-commerce brands scale profitably with data-driven PPC strategies.
Get In Touch